Food-Demand
Analysis in Mexico:
A Social Program´s Effects on
Dairy Products Expenditure
Análisis de demanda
de alimentos en México:
impacto de un programa social en el gasto
de las familias en productos lácteos
David Magaña Lemus
26
REALIDAD, DATOS Y ESPACIO. REVISTA INTERNACIONAL DE ESTADÍSTICA Y GEOGRAFÍA
This paper presents an analysis of the dairy products’
demand in Mexico at a disaggregated level. We use
data from individual household expenditure as well
as demographics from 30 169 Mexican households,
data collected in 2010 by the Mexican Statistical and
Geography Institute (INEGI). The main objective of this
study is to estimate the impact of the Oportunidades
program; which is aimed to provide financial assistance
to low-income families, whose dairy consumption is
their main source of calcium. After using the Heckman
model, we found evidence of the program beneficiaries
spending more money on some dairy products than
their counterparts. Estimates of price and expenditure
elasticities are also provided. All the dairy commodities
considered in this study are estimated to be normal
goods and necessities. The results of this paper can be
useful for public policies making in Mexico.
Key words: Heckman Model; Oportunidades program;
demand for dairy products; Mexico.
En este trabajo se presenta un análisis de demanda
de productos lácteos en México a nivel desagregado.
Se utilizan datos demográficos e información sobre el
gasto de 30 169 hogares mexicanos, información compilada en el 2010 por el Instituto Nacional de Estadística
y Geografía (INEGI). El objetivo principal del estudio es
estimar el impacto del programa Oportunidades, el cual
está enfocado a proporcionar asistencia financiera a las
familias de bajos ingresos para el consumo de productos lácteos como principal fuente de calcio. Mediante
el Modelo de Heckman se encuentra evidencia de que
las familias beneficiarias del programa gastan más en
algunos productos lácteos que sus contrapartes. Se
proporcionan estimaciones de elasticidades precio y
gasto. Todos los productos lácteos considerados en esta
investigación son estimados como bienes normales y
necesarios, de acuerdo con las elasticidades obtenidas.
Los resultados de este trabajo pueden ser útiles para la
toma de decisiones de política pública en México.
Palabras clave: Modelo de Heckman; programa Oportunidades; demanda de productos lácteos; México.
Recibido: 27 de junio de 2015
Aceptado: 11 de mayo de 2016
I. Introduction
Oportunidades program, originally known as
Progresa, is a conditional cash transfer program in
Mexico.1 Household members are required to participate in health education sessions and attend
regular health check-ups to receive the program’s
cash transfer. More than 5 million families receive
these monthly cash transfers to help them improve
their health, education, and nutrition. The health
and education components of this program are
mandatory and monitored. However, one feature
1
In 2013 the program was renamed as
Vol. 7, Núm. 2, mayo-agosto 2016.
.
of the program that is not enforced is the appropriate provision of nutrition supplements to children
(Leroy, Vermandere, Nuefeld, and Bertozzi, 2008).
Previous studies have found that Oportunidades
has a positive impact on child growth in urban
areas of Mexico (Leroy Garcia-Guerra, Garcia, and
Dominguez, 2008). Other positive effects from
Oportunidades relate to higher height growth rates
and lower incidences rates of anemia in low-income, rural infants (Rivera, Sotres-Alvarez, Habicht,
Shamah, and Villalpando, 2004).
This study is intended to provide evidence to
answer the question of whether or not receiving
27
Oportunidades translates into differential nutrition standards. Particularly, this paper will focus
on the demand of dairy products, as the main
source of calcium. The findings of Hoddinott and
Skoufias (2003) suggest that beneficiary families drink more milk than those who do not participate in the program. The objective then is to
go further and analyze how participating in the
program changes acquisition of dairy products.
Analysis on 16 disaggregated dairy products is
performed to shed light into the effect of the program on choice of food products. The relevance of
estimating the effectiveness of Oportunidades on
nutrition is evident, given the number of households that receive this benefit and for its impact
as a public policy in Mexico. As pointed out by
Pitt (1983), choice of nutrient intake in households near subsistence levels is of obvious policy
importance.
We used data from the National Income and
Expenditure Household Survey (ENIGH), which is a
nationally representative survey conducted during
2010. The analysis is performed through Heckman
two-step procedure to find the effect of the program on dairy consumption by product category.
Separability of the utility function for dairy products and for other food and non-food products was
assumed.
Using Heckman two-step procedure, the estimates of the selection equation and the demand
equation are obtained. In the selection equation, it
was found that households that receive cash transfers and training from Oportunidades have a higher
probability of consuming unpasteurized milk, unpasteurized cheese (queso fresco) and aged cheese,
compared to household with similar demographics
that did not participate in the program. Conversely,
the households that participate in the program
were found to be less likely to consume pasteurized milk, Oaxaca cheese and sour cream.
Own-price and expenditure elasticities-estimates for disaggregated dairy products are
provided in this paper. Due to the economic relevance of the dairy industry, it is expected that
28
the outcome of this paper be of interest of policy
makers and business people.
As far as the relevance of the dairy market in
Mexico is concerned, it is worth noting that the per
capita milk consumption and other dairy products
is 123 kilograms (Uribe and Torres, 2010). In 2010,
average household expenditure on dairy products
represented 12.3% of the total household expenditure on food (including expenditure on food away
from home), and 15.6% of the expenditure of food
consumed at home (INEGI, 2011). Around 60% of
the production of fluid milk is utilized in the production of dairy products, which can be referred as
industrial milk consumption. In 2010, the industrial milk consumption reached 6.9 million of metric
tons, with an average rate of growth of 2.7% between 2005 and 2010. As for the direct milk consumption, the rate of growth is 1.6% for the same
period, reaching 4.4 million tons in 2010 (Uribe
and Torres, 2011). Further, since Mexico is a net
importer of dairy products, understanding the behavior of these products in the Mexican market is
not only of interest for domestic policy makers and
local business people, but also for dairy producers
and exporters abroad. The market for dairy products in Mexico was the destination for around 20%
of the U.S. dairy exports during 2005 and 2010. It
is worth noting that the total U.S. dairy exports in
2010 were 1.6 million tons (Uribe and Torres, 2011).
This paper is organized as follows: in section
II a short summary of some of the main findings
on the effect of Oportunidades is presented. Then,
data and methods are described in sections III and
IV, respectively. Section V contains the discussion
of results and, section VI presents the main conclusions of this work as well as some opportunities for
further research.
II. Literature Review
Oportunidades (Progresa) program was created
in 1997 as a policy response to ameliorate the
economic conditions of low-income families in
México. After the 1995 crisis the level of poverty in-
REALIDAD, DATOS Y ESPACIO. REVISTA INTERNACIONAL DE ESTADÍSTICA Y GEOGRAFÍA
creased sharply. It went from 52.4% of the Mexican
population living in poverty in 1994 to 69% of
the population in 1996. Likewise, the proportion
of Mexican population living below the food poverty line increased from 21.2% in 1994 to 37.4% in
1996 (CONEVAL 2011). The program has gradually
evolved since its inception in 1997 to become what
in 2016 is known as Prospera.
A detailed exposition on the social background
where this program emerged and its evolution is
out of the scope of the paper. For an overview of
the political and economic context under which
Mexico´s Oportunidades program was introduced,
see Niño-Zarazúa (2010). Also, Czarnecki (2013)
provides a description on recent poverty measurement strategies in Mexico.
Several papers have attempted to measure
the effects of the program. Using data from the
Mexican National Health and Nutrition Survey
2006, Mundo-Rosas, Rodríguez-Ramírez, and
Shamah-Levy (2009) found evidence of a low intake of nutrients, particularly iron, zinc, folic acid,
calcium, and vitamins A and C in preschool children of low socioeconomic status, especially those
living in rural areas, and in southern Mexico. They
found the presence of nutritional problems, but
did not assess household participation in any social program aimed to reduce malnutrition.
It is also important to take into consideration
that, because of the intergenerational transmission of poor nutrition, malnutrition may have
long-term consequences. Thus, reduction in malnutrition has large potential economic benefits.
Well-nourished children have significant gains
in productivity and in cognitive achievement
(Behrman and Hoddinott, 2005).
With respect to Oportunidades, Escobar Latapí
and González de la Rocha (2008), report that a
proportion of low income families in Mexico rely
mainly on two sources of income: 1) remittances
from migration to the United States, and 2) cash
transfers from social policy programs such as
Oportunidades.
Vol. 7, Núm. 2, mayo-agosto 2016.
Since its establishment in 1997, the main objective of Oportunidades was to reach the poorest households in rural Mexico. In its initial stages,
Oportunidades stipends are allocated randomly
across localities in the evaluation sample. That is, it
was allocated to all eligible households within the
randomly selected treatment localities. Eligibility
was based on a poverty index (Behrman and
Hoddinott, 2005). In contrast to earlier programs
in Mexico, a unique feature of Oportunidades is the
targeting of transfers to the mother (Hoddinott
and Skoufias, 2003). The monthly transfer accounted for about 20% of the value of total household
consumption (Todd, Winters and Hertz, 2010).
Coverage of this program has increased significantly since its start. In June 2010 there were
5.6 million enrolled households in Oportunidades
across the country (Andalon, 2011). The program
provides cash transfers, food supplements, subsidies on school supplies, and health and education
services to families below a certain socioeconomic
threshold. In order to keep receiving the benefits,
mothers and individual beneficiaries must comply
with school attendance, health check-up as well as
attendance to health and nutrition talks (Escobar
Latapí and González de la Rocha, 2008). The program focuses on the health and nutrition of children and mothers. Leroy, Ruel and Verhofstadt
(2009) provided a useful diagram (Figure 1) that
depicts how programs such as Oportunidades may
improve children’s nutrition.
Oportunidades is a program that has been studied extensively. Some of these studies are briefly
described here. An expected effect of the nutrition
and health education component of the program,
usually targeted at women, is that households’
preferences change towards nutrient-rich foods
(Leroy, Ruel and Verhofstadt, 2009). On this matter,
Hoddinott and Skoufias (2003), found that households receiving aid from Oportunidades (they refer
to the program as Progresa), increased caloric acquisition compared to similar household not receiving the benefits. An important finding in this
study was that the higher impact was on the consumption of highly nutritious foods such as fruits,
vegetables and animal proteins.
29
Figure 1
Mechanisms by which Conditional Cash Transfer Programs might affect nutritional status
Women´s
Time
Education in Health
& Nutrition (H&N)
to Women
Health Visits
(Condition)
Education
(Condition)
Women´s
Knowledge &
Awareness
Use of H&N
Services
School
Enrolment &
Attendance
Health
Supply
HH food
Security-Diet
Quality/
Quantity
Feeding &
Care
Practices
Education
Supply
Underfying Causes
HH Income
& Women´s
Income
Control
Fortified
Products
Programme
Cash to
Women
Educated
Girls
Health
Outcomes
Child Nutrition
Long Term
Immediate
Causes
Food/
Nutrient Intake
HH refers to a household. Figure elaborated by Leroy, Ruel and Verhofstadt (2009).
Another study, conducted by Barber and Gertler
(2008) used multivariate and instrumental variable
analyses to estimate the impact of the program
on birth weight. They found that beneficiaries of
Oportunidades had children with increased birth
weights of 127.3 grams or more and a 4.6 percentage point reduction in low birth weight. It is worth
noting that pregnant women in participating
households received nutrition supplements and
health care, as well as cash transfers.
Leroy, Garcia-Guerra, Garcia, Dominguez, Rivera,
and Neufeld (2008) used difference-in-difference
propensity score matching to evaluate the impact
of Oportunidades on the growth of children younger than 24 months of age that lived in urban areas. They found that children in intervened families
and younger than 6 months of age grew more than
30
children in control families. They concluded that
Oportunidades, with its strong nutritional component, is an effective tool to improve the growth of
infants in poor urban households.
As for the program effects on a rural environment, Rivera, Sotres-Alvarez, Habicht, Shamah,
and Villalpando (2004) reported a finding of better
growth and lower rates of anemia in low-income,
rural infants and children in families that receive
Oportunidades.
Behrman and Hoddinott (2005) evaluated nutrition by estimating the impact of the program on
child height. A substantial and positive effect of
the nutritional supplements on children between
12-36 months was found. Higher growth rates and
lower probability of stunting were the main con-
REALIDAD, DATOS Y ESPACIO. REVISTA INTERNACIONAL DE ESTADÍSTICA Y GEOGRAFÍA
clusions. They found that larger effects for children
from poorer communities but whose mothers are
functionally literate.
As discussed by Andalon (2011), theoretical and
empirical health economics research suggests that
Oportunidades could have affected adolescent
weight. This author conducted a study on the causal effect of the program on overweight and obesity rates of youth living in rural areas. He found
a reduction in the rate of obesity among female
adolescents who participated in the program for
4 years on average. This effect is attributed by the
author to increased schooling and access to information regarding the consequences of unhealthy
behavior through health information sessions and
regular visits to physicians; the education and nutrition component of the program. Further, Andalon
(2011) suggested that given the identified local average treatment effect (LATE) at the threshold for
program eligibility, female obesity would decrease
if the program was expanded to cover households
with a slightly better economic status.
On the impact of Oportunidades on productive
activities, Todd, Winters and Hertz (2010), using first
difference and weighted estimators (weights were
constructed from propensity scores), evaluated the
impact of the program on food consumption from
own production, land use, livestock ownership,
and agricultural spending. The results they obtained support the hypothesis that transfers influence agricultural production. Greater impacts were
found for households invested in agriculture.
As it is clear from the cited literature, there is
plenty of evidence that suggest that Oportunidades
program has been successful in helping poor families to achieve better nutrition levels. However, to
the best of my knowledge no study has been conducted, using the ENIGH, to explore at a disaggregated level the type of products that account for
the increased intake of nutrients. In this paper I will
focus on the analysis of demand for dairy products
as the main source of calcium (an important micronutrient). Moreover, the objective of this study
is to assess the impact of Oportunidades on the
Vol. 7, Núm. 2, mayo-agosto 2016.
probability of purchasing these types of products
(dairy) as a way to shed light into the factors that
impact the decision making process to purchase or
not products that contain calcium. The analysis is
conducted utilizing the Heckman two-step estimator, in which the selection equation, that measures
the probability of purchasing dairy products or
not, and the demand equation, that estimates the
quantity consumed once the household decided
to purchase dairy products, are specified.
III. Data
The data used in this study was obtained from
the National Income and Expenditure Household
Survey (ENIGH), which is a nationally representative
income and expenditure survey that was conducted
by the Mexican Statistical and Geography Institute
(INEGI) during the third quarter of 2010. This survey
contains data on food consumption (at home and
away from home) and the money value of food
consumed during one week for 30 169 households.
The data set also contains socioeconomic characteristics of the household, including geographical
location, housing characteristics, number and age
of members of the family, income, gender, and education level of the head of the household, among
other factors. Moreover, the survey contains information about the marginality index of every county.
This is a measure of the lack of access to elementary
education, health services, infrastructure and development opportunities, which translates into low income families living in poor housing conditions. In
the survey, all the sources of household income are
disaggregated, allowing the identification of households that are beneficiaries of social programs such
as Oportunidades.
As for the data manipulation to obtain the variables that were used in the model, and to obtain
an estimation of prices, the household expenditure
of a specific commodity was divided by the quantity they purchased. As expected in a household
survey, it is common to find that a household does
not report expenditure for all the commodities of
interest. In those cases it is not possible to calcu-
31
late a price for that specific commodity and household. To avoid missing values, the average price
of the State was imputed for the commodity not
consumed by the household. Expenditure on dairy
was obtained by merely adding up the household
expenditure for the 16 commodities under analysis. In this way, values for the variables quantity and
price for the 16 dairy products, as well as for total
expenditure on dairy were available for analysis. It
is worth noting that the focus of this paper is on
the food expenditure at home.
To account for the proportion of children living
in a house, the number of children under 12 years
old is divided by the total number of persons in the
household. This variable is labeled as children ratio.
The rest of the independent variables in the
model are as follows:
Marginalization index, which has 5 levels, is operationalized as a set of dummy variables, with value equals to 1 if the household belongs to a given
level of marginalization, and equal to zero otherwise. The marginalization levels, according to the
Mexican Agency known as CONAPO, are: Very High,
High, Medium, Low and Very Low. In the model the
level Very High marginalization is the reference level, that is, those households that live in the most
unfavorable conditions. The variable female is a
dummy variable that stands for the gender of the
head of the household. This is equal to 1 if the head
is a female, and zero otherwise. Age is an interval
variable that measures the age of the head of the
household. The variable Oportunidades is a dummy
variable, which is equal to 1 if the family reported
income from the program in 2010, zero otherwise.
The ENIGH 2010 collected data on 18 different
dairy categories. However, for two of them, almost
all households reported zero consumption during
the survey period. For this paper, 16 categories of
dairy products are considered. These are: pasteurized milk, condensed milk, evaporated milk, powder milk, modified milk or formula, unpasteurized
milk, cheddar cheese, aged cheese, unpasteurized
(fresh) cheese, manchego cheese, Oaxaca cheese,
32
other types of cheese, sour cream, butter, yogurt,
and other dairy products.
IV. Methodology
Estimation at the household level is preferred because it more readily permits the incorporation
of household composition variables into the demand analysis, such as household size, occupation
and employment status—information that is typically lost in aggregation. This is of particular importance in the study of nutritional well-being as it is
the poorest households which are of special interest (Pitt, 1983).
As stated in the literature, the use of householdlevel data is preferable, since it avoids the problem
of aggregation over consumers and often provides
a large and statistically rich sample. However, data
based on this kind of surveys convey a major estimation problem, namely zero consumption for many
goods in a given household. Hence, we would be
facing a censored dependent variable (Hein and
Wessells, 1990).
As pointed out by Keen (1986), the presence of
zero consumption values can be because of any
of three broad factors:
i) Infrequent purchasing (household budget
surveys commonly record expenditures
only over a relatively short period of time).
ii) Variation in preferences (and/or prices)
across the sample (households may simply
not consume some commodities);
iii) Misreporting.
When for at least one commodity there are
households with zero consumption level, the application of the usual continuous variable estimation
techniques would result in biased and inconsistent
estimates because the random disturbances have
nonzero means and are correlated with the exogenous variables. Moreover, dropping those households which do not consume at least one of the
goods would severely reduce sample size and still
REALIDAD, DATOS Y ESPACIO. REVISTA INTERNACIONAL DE ESTADÍSTICA Y GEOGRAFÍA
result in inconsistent estimates (Pitt, 1983). To deal
with this estimation problem some techniques
have been proposed over the years, including the
widely-used Tobit analysis, originally proposed by
Tobin (1958). However, the Tobit regression model
assumes that the probability mechanism generates both the zeros and the positive values in the
dependent variable. This is a strong assumption
(Cameron and Trivedi, 2010). If zero consumption is
assumed to be due to sample selection, Heckman’s
two-step is the appropriate model (Chern, Ishibashi
and Taniguchi, 2003).
As discussed in Cameron and Trivedi (2010), the
two-part model, unlike the Tobit model, attains
some of its flexibility by assuming that: 1) the decision to spend, and 2) the amount to spend, are
independent. That is, it has two-parts that are independently determined. The two-equation model
includes a selection equation for , and an outcome equation for .
, and
where
is a latent variable that represents the
outcome of interest, quantity purchased in our
case. There is also the latent variable . The out. In other words,
come of is observed only if
determines whether the household buys dairy
products (can also be interpreted as the propendetermines the quantity
sity to consume) and
demanded. The regression model is:
where and are the error terms assumed to be
uncorrelated with the covariates, but may be correlated between them. It is assumed that the
correlated error terms are jointly normally distribare
uted and homoskedastic. Moreover, and
vector of covariates for the selection and the outand
come (demand) equation, respectively
are vectors of coefficients for the selection equation (Probit model) and the demand equation (linear model).
Vol. 7, Núm. 2, mayo-agosto 2016.
It is important to note that a two-step method
is based on the following conditional expectation:
where
given,
and the normality of the errors, then
. The term ρ measures the correlation between the error terms and . The term
can be estimated by
and
is obtained by a Probit regression of on
. As for the least squares part, it is estimated by
regressing on in the second step. The generated regressor, called the inverse Mills’ ratio is
, it is also known as the non-selection
hazard (Cameron and Trivedi, 2010). This term
takes into account a possible selection bias.
For analysis, the marginal effects of independent variables were estimated. Special interest lies
on the effect of the dummy variable Oportunidades on the consumption of dairy products. As discussed by Greene (2003), the marginal effect of
the independent variables on in the observed
sample consists of two components. First, there is
the direct effect of the independent variable on the
which is captured by the coefficients.
mean of
Second, if the independent variable also appears
in the selection equation, there is an indirect effect. This is because a change in some independent variables not only changes the mean of ,
but also the probability that an observation is in
the sample. Thus, the independent variables will
affect through the lambda term.
As for identification, the Heckman model is
identified when the same independent variables
in the selection equation appear in the outcome
equation. However, Heckman model is appropriate
when at least one explanatory variable is included
in the selection equation, but not in the outcome
equation. This is especially problematic when having small sample sizes because the Heckman procedure yields inconsistent and biased results when
33
the exclusion restrictions are missing (Sartori,
2003). Also, Woldrigde (2010) points out that if all
the variables in the selection equation also appear
in the demand equation, the Heckman two-step
(Heckit) estimates become imprecise.
To avoid this problem, in this case the variable
that was included in the selection equation but
not in the demand equation was the total quarterly income of the household. The dairy expenditure,
a variable needed in the demand equation, was
calculated as the sum of the expenditures for each
dairy product.
V. Results and Discussion
Let us start this section with some descriptive
statistics on the consumption of dairy products.
Table 1 provides the mean and the standard deviation of quarterly consumption of dairy products
by household. At household level, data is divided
to compare the consumption level of families that
receive Oportunidades and those that do not. For
most of the 16 dairy products the average quarterly consumption is higher for those households
that did not participate in the program. This is not
surprising since dairy products are normal goods
and households that received cash transfers from
the Mexican government, through the social program, are low income families. On the dispersion
Table 2
Table 1
Descriptive statistics of dairy products
purchased by category for Oportunidades
beneficiaries and nor beneficiaries in Mexico
Variable
Households with
Oportunidades Households without
Oportunidades
Mean Std. dev. Mean Std. dev.
Pasteurized milk
22.03
35.99
41.24
50.06
Condensed milk
0.04
0.50
0.15
2.17
Evaporated milk
0.02
0.37
0.16
2.17
Powder milk
1.02
5.99
0.73
5.94
Modified milk or
formula
0.10
2.85
0.10
1.72
Unpasteurized milk
5.28
32.29
3.20
20.63
Cheddar cheese
0.06
1.39
0.17
1.39
Aged cheese
0.37
1.73
0.18
1.11
Unpasteurized
(unaged) cheese
2.14
4.26
1.63
3.78
Manchego cheese
0.12
0.71
0.25
1.26
Oaxaca cheese
0.46
2.40
1.00
2.85
Other cheese
0.45
1.75
0.75
2.71
Sour cream
0.67
2.22
1.22
2.86
Butter
0.10
0.64
0.21
1.12
Yogurt
1.27
4.34
2.48
7.39
Other dairy
0.22
1.76
0.34
2.18
Quantities are measures in liters/kilograms per household for a three-month
Descriptive statistics of demographic variables for Oportunidades beneficiaries
and not beneficiaries in Mexico
Variable
Households with Oportunidades
Mean Std. Dev. Households without Oportunidades
Mean Std. Dev.
Age of the head of the household
47.31
15.29
48.01
15.45
Total household members (1)
4.94
2.20
3.76
1.80
Children under 12 years (2)
1.42
1.34
0.83
1.03
Ratiokids [(1)/(2)]
0.25
0.22
0.18
0.21
Quarterly household income
6 118.72
9 087.03
16 580.82
23 103.79
Quarterly expenditure on dairy
651.49
657.05
953.15
909.15
Income and expenditure are measured in 2010 Mexican pesos.
34
REALIDAD, DATOS Y ESPACIO. REVISTA INTERNACIONAL DE ESTADÍSTICA Y GEOGRAFÍA
of the consumption distribution, most of the commodities have a high standard deviation. It implies
a highly skewed distribution to the right (since the
lower bound of the distribution is zero).
With respect to the descriptive statistics of the
demographic variables used in the model, the head
of the household is a female for 22% of the households receiving Oportunidades, compared to 26%
in the households that are not part of the program.
The mean and standard deviation for variables
such as age of the head of the household, total
members of the household, number of children
under 12 years old, as well as quarterly income and
expenditure on dairy is reported in Table 2.
As for the empirical results, the Heckman model described in the previous section was estimated equation by equation for each of the 16 dairy
products under analysis. The main objective was to
evaluate the effect of participating in the program
Oportunidades on the probability that the household purchased a given dairy product, controlling
for the log of prices for all the dairy products, and
the log of expenditure on dairy and demographic
characteristics. In particular, the age and gender of
the head of the household, the marginality index
(as described in section III), as well as the ratio of
kids under 12 years old to total number of persons
that live in the household were controlled for in
every equation. As mentioned by Cameron and
Trivedi (2010) the two-part model uses a dependent variable in logs.
The first panel in Table 3 contains the coefficients
of the first step in the Heckman model. The signs of
the coefficients that are statistically significant in the
probit part of the model suggest that Oportunidades
has a positive impact on the probability of buying
unpasteurized milk (p<.01), aged cheese (p<.05)
and unpasteurized unaged cheese (p<.01). On the
other hand, the probit coefficients for the variable
Oportunidades have a negative sign for the equations for which the dependent variable is pasteurized milk (p<.01), Oaxaca cheese (p<.05) and sour
cream (p<.05). For the rest of the products the coefficients are not statistically significant.
Vol. 7, Núm. 2, mayo-agosto 2016.
Table 3
Coefficients of the variable Oportunidades
from the Probit model (first step)
and the marginal effects (two-step)
Estimated equation by equation
Two-step
Average Marginal
Effect
Dependent variable Probit
(First step) Pasteurized milk
-0.17 **
-0.049 **
(0.03)
0.010
Condensed milk
-0.03
0.001
(0.09)
0.003
Evaporated milk
-0.23
-0.007
(0.13)
0.004
Modified milk or
formula
0.08
0.001
(0.14)
0.002
0.18 **
0.016 **
(0.05)
0.005
Cheddar cheese
-0.01
-0.001
(0.06)
0.005
Aged cheese
0.14 *
0.009 *
(0.06)
0.004
0.11 **
0.033 **
(0.03)
0.010
Manchego cheese
-0.05
-0.004
(0.07)
0.005
Oaxaca cheese
-0.10
-0.023
(0.04)
0.010
Other cheese
0.04
0.007
(0.04)
0.009
Sour cream
-0.11 *
-0.030 *
(0.04)
0.010
Butter
0.01
0.001
(0.06)
0.006
Yogurt
-0.02
-0.004
(0.04)
0.010
Other dairy
-0.03
-0.002
(0.06)
0.005
Unpasteurized milk
Unpasteurized
(unaged) cheese
Standard errors in parentesis.
Delta method standard errors for two-step estimates are reported below its
coefficient.
35
From the conditional expectation equation presented in section IV, it is easy to see that if ρ ≠ 0 and if
the independent variable appears in both equations
(selection and demand), then the Oportunidades coefficients do not indicate the marginal effect of the
program on the probability of purchasing a given
dairy product. As mentioned by Greene (2003), it is
possible that the sign, magnitude and statistical significance of the marginal effect differ to those in the
beta coefficients.
The ρ term was not statistically different from zero,
i.e. the error terms in the selection equation and in the
demand equation were uncorrelated, for the equations that had modified milk or formula, unpasteurized milk, butter and other dairy as the dependent
variable. As mentioned by Wooldridge (2010), a
standard t test on ρ is a valid test of the null hypothesis of no selection bias. However, for the eleven
dairy commodities under analysis ρ was significantly different from zero (p<.05), hence in these cases
the selection process and the demand equation are
not uncorrelated. In the case of powder milk, the
probit coefficient and the marginal effect were not
estimated since the dependent variable was never
censored because of selection, thus the model simplified to ordinary least squares regression.
The second panel of Table 3 introduces the average marginal effects of Oportunidades on the
probability of purchasing dairy products. In particular, according to these estimates, a household
that receives the benefits of the program increases the probability of consuming: 1) unpasteurized milk in around 2 percentage points; 2)
aged cheese in around 1 percentage point; and
unpasteurized cheese, also known as queso fresco, in around 3 percentage points. These average
marginal effects are comparisons with respect to a
household, of similar demographics, that is not a
beneficiary of the program.
On the other hand, there is evidence that suggest that a household receiving Oportunidades is
less likely to have a positive expenditure on (i.e.
purchase) pasteurized milk (around 5 percentage
points less likely), Oaxaca cheese (around 2.3 per-
36
centage points less likely) and sour cream (3 percentage points less likely) compared to a household that
does not participate in the program.
If we consider that a decrease in expenditure
does not necessarily mean a decrease in consumption (Aguiar and Hurst, 2005), then the decrease
on probability of purchasing a positive quantity of
pasteurized milk, Oaxaca cheese and sour cream,
does not necessarily imply that Oportunidades has
a negative impact on dairy consumption.
In this sense, the decrease in the probability of
purchasing pasteurized milk may be explained by
a substitution towards purchasing unpasteurized
milk, which is produced locally in rural areas, or
towards own-production. The nutrition talks, part
of the program, might had an effect by providing
training in managing unpasteurized milk (boiling
it) to avoid the risk of food borne disease.
With respect to the decrease in the probabilities of purchasing Oaxaca cheese, it may be the
case that it is produced at home, since it is relatively easy to elaborate manually. This is not an unfounded conclusion if we consider the findings of
Todd, Winters and Hertz (2010), who reported that
Oportunidades increased the value and variety of
food consumed from own-production. They also
reported an increase in livestock ownership, a fact
that could explain the decline in the probability of
purchasing pasteurized milk and Oaxaca cheese.
Elasticities
The estimated elasticities are own-price and expenditure. The later refers to the expenditure on dairy
products, which was estimated under the separability assumption. With respect to the own-price
elasticity estimates, it was found that all the parameters had a negative sign, as expected, and all
but one of them are statistically significant (p<.01).
The products that are estimated to be inelastic are
pasteurized milk, powder milk, formula, unpasteurized milk, unpasteurized cheese, manchego
cheese, sour cream, butter, and the category for
REALIDAD, DATOS Y ESPACIO. REVISTA INTERNACIONAL DE ESTADÍSTICA Y GEOGRAFÍA
other dairy products. Also, as shown in the second
panel of Table 4, dairy products such as cheddar
cheese, Oaxaca cheese, other cheese and yogurt
are estimated to be own-price elastic. Moreover,
condensed milk and evaporated milk are highly
own-price elastic products (-2.25 and -2.33, respectively). Own-price elasticity for aged cheese was
not statistically significant. As for the expenditure
elasticity estimates, all dairy products have a positive and significant parameter (p<.01). Some dairy
products, specially the different types of milk, have
a large expenditure elasticity. However, in all cases
the expenditure elasticities are less than unity.
These findings are consistent with the results reported by Skoufias, Di Maro, Gonzalez-Cossio, and
Rodriguez Ramirez (2009). They estimated, using a
sample of poor rural household in Mexico, income
elasticity for a variety of macro and micronutrients,
including calcium. The income elasticities for calcium, estimated by Skoufias, Di Maro, GonzalezCossio, and Rodriguez Ramirez (2009), go from
0.6 to 0.99 (p<.01) depending on the method used.
Large income elasticities for dairy products imply
that an increase in household disposable income,
which can be achieved by cash transfers and/or by
increased productivity, would be an effective way
to increase consumption of basic nutrients, such as
calcium, in poor households in Mexico.
VI. Conclusions
Using Heckman two-step procedure, the estimates
of the selection equation and the demand equation were obtained. In the selection equation, it
was found that households that receive cash transfers and training from Oportunidades had a higher
probability of consuming unpasteurized milk, unpasteurized cheese (queso fresco) and aged cheese,
compared to household with similar demographics that did not participate in the program.
Conversely, the households that participate in the
program were found to be less likely to consume
pasteurized milk, Oaxaca cheese and sour cream.
These results suggest a substitution effect towards
Vol. 7, Núm. 2, mayo-agosto 2016.
Table 4
Expenditure and own price elasticities
for dairy products in Mexico
Estimated by Heckman´s two-step procedure,
equation by equation
Dependent
variable
Expenditure
elasticity
(wrt dairy
expenditure) Own price elasticity Pasteurized milk
Condensed milk
Evaporated milk
Powder milk
0.75 **
(0.01)
0.67 **
(0.06)
0.82 **
(0.09)
0.85 **
(0.03)
-0.85 **
(0.02)
-2.25 **
(0.18)
-2.33 **
(0.27)
-0.90 **
(0.03)
Modified milk or
formula
0.99 **
-0.99 **
Unpasteurized milk
Cheddar cheese
Aged cheese
(0.26)
0.68 **
(0.04)
0.40 **
(0.03)
0.38 **
(0.03)
(0.26)
-0.87 **
(0.13)
-1.49 **
(0.11)
-0.07
(0.14)
Unpasteurized
(unaged) cheese
0.46 **
-0.97 **
Manchego cheese
Oaxaca cheese
Other cheese
Sour cream
Butter
Yogurt
Other dairy
(0.01)
0.24 **
(0.03)
0.56 **
(0.02)
0.47 **
(0.02)
0.40 **
(0.01)
0.12 **
(0.03)
0.56 **
(0.03)
0.15 **
(0.05)
(0.04)
-0.48 **
(0.09)
-1.36 **
(0.09)
-1.24 **
(0.09)
-0.92 **
(0.14)
-0.76 **
(0.10)
-1.49 **
(0.11)
-0.42 **
(0.19)
Standard errors in parentesis.
*p<.05, **p<.01
37
products that have a component of household
production, such as boiling raw milk and production of cheese.
Own-price and expenditure elasticities estimates for disaggregated dairy products are also
provided in this paper. Due to the economic relevance of the dairy industry in Mexico, it is expected that the outcome of this paper be of interest of
policy makers and business people.
As stated in the paper, previous studies have
suggested an increase in milk consumption as the
impact of receiving the benefits of the program. In
this paper, the effect on different dairy products
expenditure was obtained. However, the overall
effect of Oportunidades on calcium consumption
by household was not estimated. Opportunities
for further research include the estimation of consumption of dairy products accounting for calcium
content by individual members of a household.
Likewise, further analyses can be conducted controlling for different demographic variables, such as
the level of education of the head of the family, the
number of individuals in the family that report income, the occupation of families (agricultural producers or not), geographical region, whether the
family belongs to an indigenous community, etc.
Given the disaggregation of the data set, home
production for self-consumption is not distinguished in this paper (which is a subject for further
research). Also, separability was assumed but not
formally tested. This paper focuses on milk consumption as indicator of calcium consumption.
Nevertheless, recommended intake of calcium was
not considered in the analysis. Certainly, further effort is needed to complement this study with other nutrients and their demand analysis. Moreover,
the analysis is done using only data from INEGI,
more work needs to be done using different data
source to test if the results are robust. One major
area of opportunity for further research is to estimate the equations as a system, instead of equation by equation, in order to be able to estimate
cross-price elasticities and formally test adding
up and symmetry restrictions in the demand sys-
38
tem. Finally, in order to improve robustness of the
findings it would be interesting to perform estimation of the impact of the program on nutrition in
low-income households using two-equation joint
estimation and a Bayesian approach.
References
Aguiar, M. and Hurst E. 2005. “Consumption Versus Expenditure.” Journal of
Political Economy, 113(5): 919-48.
Andalon, M. 2011. “Oportunidades to Reduce Overweight and Obesity in
Mexico?”. Health Economics, 20 (Suppl. 1): 1-18 (2011).
Barber, S. L., and P. J. Gertler. 2008. “The Impact of Mexico’s Conditional Cash
Transfer Programme, Oportunidades, on Birthweight”. Tropical Medicine
and International Health, volume 13 No. 11 pp 1405-1414.
Behrman, J. R. and J. Hoddinott. 2005. “Programme Evaluation with
Unobserved Heterogeneity and Selective Implementation: The Mexican
PROGRESA Impact on Child Nutrition”. Oxford Bulletin of Economics and
Statistics, 67, 4 (2005) 0305-9049.
Cameron, A. C. and P. K. Trivedi. 2010. “Microeconometrics Using Stata”.
Revised edition. Stata Press, College Station, Texas.
Chern, W. S., K. Ishibashi, and K. Taniguchi. 2003. “Analysis of the Food
Consumption of Japanese Households”. FAO, Social and Economic
Development Paper No. 152. Rome, Italy, 2003.
CONEVAL. 2011. “Medición de la pobreza en México, 2008-2010”. Available at
http://www.coneval.gob.mx/cmsconeval/rw/pages/medicion/Pobreza
_2010/Anexo_estadistico.es.do (Accessed on September 13th, 2015).
Czarnecki, L. 2013. “El CONEVAL. La institucionalización del concepto oficial
de la pobreza en México”. Revista Castellano-Manchega de Ciencias
Sociales, Nº 16, pp. 177-190.
Escobar Latapí, A. and M. González de la Rocha. 2008. “Evaluating
Progresa-Oportunidades”.
In S. Razavi (ed.) The Gendered Impacts of Liberalization: Towards Embedded
Liberalism? Chapter 10, pp:435-468. Routledge/UNRISD, New York, 2008.
Greene, W. H. 2003. Econometric Analysis. New Jersey: Prentice Hall.
Hoddinott, J. and E. Skoufias, Y. 2003. “The Impact of PROGRESA on Food
Consumption”. FCND Discussion Paper No. 150, International Food Policy
Research Institute, Washington, DC.
Hein, D and C. R. Wessells. 1990. “Demand Systems Estimation with
Microdata: A Censored Regression Approach”. Journal of Business &
Economics Statistics. Vol. 8, No. 3 (Jul., 1990), pp. 365-371.
Heckman, James. 1979. “Sample Selection Bias as a Specification Error.”
Econometrica 47:153-161.
INEGI. 2011. Encuesta Nacional de Ingresos y Gastos de los Hogares 2010.
Instituto Nacional de Estadística y Geografía, July 2011. Accessed on
REALIDAD, DATOS Y ESPACIO. REVISTA INTERNACIONAL DE ESTADÍSTICA Y GEOGRAFÍA
April 2014. Available at http://www.inegi.org.mx/est/contenidos/
Proyectos/Encuestas/Hogares/regulares/Enigh/Enigh2010/tradicional/
default.aspx
Keen, M. 1986 “Zero Expenditures and the Estimation of Engel Curves”.
Journal of Applied Econometrics, 1, 277-286.
Leroy, J. L., H. Vermandere, L. M. Neufeld, and S. M. Bertozzi. 2008.
“Improving Enrollment and Utilization of the Oportunidades Program
in Mexico Could Increase Its Effectiveness”. The Journal of Nutrition, 638641, 2008.
Leroy, J. L., A. Garcia-Guerra, R. Garcia, C. Dominguez, J. Rivera, and L. M.
Neufeld. 2008. “The Oportunidades Program Increases the Linear Growth
of Children Enrolled at Young Ages in Urban Mexico”. The Journal of
Nutrition, 138: 793-798, 2008.
Leroy, J. L., M. Ruel, and E. Verhofstadt. 2009. “The Impact of Conditional
Cash Transfer Programmes on Child Nutrition: a Review of Evidence using
a Programme Theory Framework”. Journal of Development Effectiveness,
Vol. 1, No. 2, June 2009, 103-129.
Leroy, J. L., P. Gadsden, S. Rodríguez-Ramírez, and T. Gonzalez de Cossio.
2010. “Cash and In-Kind Transfers in Poor Rural Communities in Mexico
Increase Household Fruit, Vegetable, and Micronutrient Consumption
but Also Lead to Excess Energy Consumption”. The Journal of Nutrition,
140: 612–617, 2010.
Mundo-Rosas, V., S. Rodríguez-Ramírez, and T. Shamah-Levy. 2009. “Energy
and Nutrient Intake in Mexican Children 1 to 4 years old. Results from the
Mexican National Health and Nutrition Survey 2006”. Salud Pública Méx.,
2009;51 suppl 4:S530-S539.
Niño-Zarazúa, M. 2010. “Mexico’s Progresa-Oportunidades and the
emergence of social assistance in Latin America”. Brooks World Poverty
Institute, BWPI Working Paper 142. ISBN : 978-1-907247-41-5.
Vol. 7, Núm. 2, mayo-agosto 2016.
Pitt, M. M. 1983. “Food Preferences and Nutrition in Rural Bangladesh”. The
Review of Economics and Statistics, 65, 105-114 pp.
Rivera, J.A., Sotres-Alvarez, D., Habicht, J.P., Shamah, T., Villalpando, S. 2004.
“Impact of the Mexican program for education, health and nutrition
(Progresa) on rates of growth and anemia in infants and young children:
a randomized effectiveness study”. JAMA. 2004; 291:2563-70.
Sartori, A. 2003. “An Estimator for Some Binary-Outcome Selection Models
Without Exclusion Restrictions.” Political Analysis 11:111-138.
Skoufias, E., V. Di Maro, T. Gonzalez-Cossio, S. Rodriguez Ramirez. 2009.
“Nutrient Consumption and Household Income in Rural Mexico”.
Agricultural Economics, 40 (2009) 657-675.
Tobin, J. 1958. “Estimation of Relationships for Limited Dependent Variables”.
Econometrica 26, 24-36.
Todd, J. E., P. C. Winters and T. Hertz. 2010. “Conditional Cash Transfers and
Agricultural Production: Lessons from the Oportunidades Experience in
Mexico”. Journal of Development Studies, 46:1, 39-67.
Uribe, L., and E. Torres. 2010. “Panorama Agroalimentario Leche y Lácteos
2010”. FIRA-Banco de México. Accessed March 2014. Available at http://
www.fira.gob.mx
Uribe, L., and E. Torres. 2011. “Panorama Agroalimentario Leche y Lácteos
2011”. FIRA-Banco de México. Accessed March 2014. Available at http://
www.fira.gob.mx
Wooldridge, J. 2010. Econometric Analysis of Cross Section and Panel Data.
Cambridge: MIT Press. Second edition.
39